Apple Maps sharpens 3D city views

Apple Maps is preparing a major upgrade to its three-dimensional city experience, with 3D Gaussian splatting set to give users cleaner, sharper and more realistic urban scenes built from oblique aerial imagery.

The shift marks a technical break from conventional mesh-based photogrammetry, the approach that has long powered digital city models but often produced warped tree canopies, distorted cables, softened building edges and streetscapes that appeared convincing only from a distance. Gaussian splatting uses dense collections of semi-transparent 3D points, or “splats”, to reproduce shape, colour, light and depth with higher visual fidelity, allowing complex details to hold together from multiple viewing angles.

For Apple Maps, the change could strengthen one of the platform’s most visually distinctive features: immersive 3D city navigation. The company has spent years expanding detailed city experiences, Look Around street-level views and Flyover-style aerial perspectives across major urban centres. A move towards splat-based rendering would give the service a more photorealistic layer, especially in areas where traditional reconstruction struggles with fine structures such as trees, overhead wires, roof edges, balconies, traffic signs and irregular street furniture.

The scenes are understood to be generated from oblique aerial imagery, captured at angles rather than directly from above. That matters because angled image sets provide more information about façades, street corridors, rooflines and vertical surfaces. When processed through Gaussian splatting methods, those images can produce scenes that preserve ground-level detail without relying entirely on the rigid triangle meshes used in older 3D mapping pipelines.

The technology has gained rapid traction across computer vision, geospatial mapping, digital twins and spatial computing because it can render photorealistic 3D environments in real time. Instead of converting imagery into a conventional polygon model, Gaussian splatting represents a scene through millions of small, soft volumetric elements. Each carries data about position, scale, orientation, colour and transparency, allowing a viewer to move through a scene while the system blends these elements into a coherent image.

Apple’s interest fits a wider strategy across mapping, spatial media and Vision Pro. Its software ecosystem is increasingly built around immersive visual experiences, with developers encouraged to create richer 3D environments for apps, websites and spatial computing workflows. Better Maps data also supports navigation, tourism, property search, gaming, urban planning and augmented reality applications, where realistic geometry and visual stability are essential.

The most immediate consumer impact will be visual quality. Older photogrammetry often looked impressive at city scale but broke down on inspection. Trees could turn into indistinct green clumps, bridges could smear at their edges, and thin infrastructure could appear bent or melted. Gaussian splatting is better suited to preserving such visual complexity, particularly when source imagery has strong multi-angle coverage. For users, that could mean smoother city previews, more recognisable landmarks, cleaner neighbourhood views and a stronger sense of place before arriving at a destination.

The competitive implications are also significant. Google Maps and Google Earth remain the global benchmark in mapping coverage, data volume and street-level imagery. Apple, however, has been investing steadily in higher-quality cartography, privacy-focused navigation and curated city experiences. A sharper 3D layer would not erase Google’s scale advantage, but it could help Apple compete on visual quality in supported locations and deepen the appeal of Maps across iPhone, iPad, Mac and Vision Pro.

The rollout is likely to be selective at first. High-quality splat-based city models require dense, well-aligned imagery, careful processing and efficient compression for delivery to consumer devices. Dense urban cores, landmarks, tourism districts and cities already covered by Apple’s detailed 3D experiences are the most likely early candidates. Broader coverage would depend on aerial capture cycles, licensing arrangements, local regulation and the cost of processing vast quantities of imagery.

There are also unresolved questions. Photorealistic city models raise privacy and security sensitivities, particularly around sensitive facilities, residential areas and real-time accuracy. Mapping companies typically blur faces and licence plates in street-level imagery, while aerial and 3D datasets may require additional safeguards. More realistic scenes could also increase pressure on platforms to keep visual data current, because outdated building layouts or construction zones may become more noticeable when rendered with greater fidelity.